Loading...
Loading...
Stop shipping RAG demos that forget the previous turn. Build a conversational tutor over a full book with chapter-aware chunking, local embeddings, a persistent FAISS index, and memory that actually follows the thread.
Message a mentor about fit, prerequisites, or where to start. Replies come on WhatsApp, usually within a day.
Engineers are learning here from
Turn any PDF book into an AI tutor with chapter-aware chunking, semantic retrieval, and multi-turn conversational memory. Build a document QA system that handles follow-up questions without losing context.
Build a conversational AI tutor over long PDFs with chapter-aware chunking and memory-backed retrieval.
What you'll ship
What you'll learn
Curriculum
Long-doc RAG vs chat RAG
Understand why adding chat history to retrieval changes the problem, and see the shape of a conversational retrieval chain
Docling loader
Parse long PDFs with OCR and table structure, convert to markdown with chapter headings, and chunk for retrieval
Local embeddings and FAISS
Embed chunks locally with HuggingFace sentence-transformers, build a FAISS index, persist it to disk, and reload fast
ConversationalRetrievalChain
Wire the chain that rewrites follow-ups, retrieves grounded chunks, and refuses to hallucinate
Interactive tutor
Wrap the chain in a REPL, handle follow-ups across turns, and finish the workshop with a final checkpoint
Who it's for
who shipped a basic RAG demo and watched it break on the second question in a conversation
working with long technical documents, textbooks, or manuals that need chapter context to answer well
who want a local PDF tutor that actually remembers what was asked earlier in the session
FAQ
No. The workshop uses sentence-transformers/all-MiniLM-L6-v2, which runs well on CPU. You only need an OpenRouter or OpenAI key for the chat model.
Docling preserves layout, runs OCR on scanned pages, and extracts table structure. For long books and technical PDFs, that metadata is the difference between clean chunks and garbled text.
For a single-book tutor that runs locally, FAISS is ideal. It is fast, has zero infrastructure, and persists to disk. For multi-tenant production systems, move to Qdrant or pgvector afterward.
No. The workshop focuses on ConversationalRetrievalChain and the retrieval prompt. You will walk away with a clear mental model of how memory, retrieval, and grounding fit together.
Pricing
Subscribe to Pro for every paid course, or buy just this one.
Unlock this course and every paid course plus workshop replays. One subscription.
You save 54% with regional pricing
One-time purchase. Lifetime access to every lesson, exercise, and update.
You save 41% with regional pricing
Still deciding? Ask Param a question
Conversational retrieval is the baseline for every serious tutor. Start here.
Long Document RAG with Conversational Memory
$29 one-time